To improve the impact resistance of composite materials, this study introduces an efficient global optimization framework based on a multi‐fidelity surrogate model to optimize the layup sequence of laminates. Numerical simulation models are constructed utilizing Hashin and Puck failure criteria to provide a few time‐consuming high‐fidelity samples and a multitude of inexpensive low‐fidelity samples. The traditional genetic algorithm has been improved in terms of encoding mechanisms, genetic operators, genetic probabilities, algorithm structure, and the recognition of constraint conditions. The translation propagation algorithm based on optimal Latin hypercube sampling is used to generate initial sample points and improve the quality of space filling. An efficient global optimization framework for time‐consuming simulation problems is established by combining the Hierarchical Kriging surrogate model and multi‐fidelity expectation improved addition criterion to seek the optimal layup sequence of composites when the impact resistance of unconventional layup laminates is maximized. The optimized laminate shows a significant increase in impact resistance, evidenced by a 32.5% increase in peak impact contact force and an 18.01% decrease in the total delamination damage area. Analysis of the impact damage area reveals that post‐optimization, the laminate deforms less under impact load, exhibiting more directions of damage spread and an enhanced load‐bearing capacity.Highlights
An improved DNA genetic algorithm is proposed in terms of encoding mechanisms, genetic operators, and algorithm structure.
Numerical simulation models are constructed utilizing Hashin and Puck failure criteria to provide high‐fidelity samples.
An efficient global optimization framework for time‐consuming simulation problems is established by combining the Hierarchical Kriging surrogate model.
The mechanical response and damage mechanism of the original and optimized laminates under low‐velocity impact are analyzed.